Home / ubuntu / Questions / 1053023
Accepted
Socrates
Asked: 2018-07-08 08:10:54 +0800 CST

Measure every single command within a bash file?

  • 772

Is there a way to measure the time spent in every single command within a bash file? Hence, put time in front of every command. The quantity of commands is unknown right now, as I intend to use this time measure for future bash scripts, too. And also worth mentioning is that I intend to only run simple bash scripts having command1, command2, command3, ... etc. Hence, no complex higher logic scripts.

Lets say I have a bash file similar to this one:

#!/bin/bash

mv /path/to/file /other/path/
cd /other/path/
tar -xzf /other/path/file

Is there a way to get some output similar to this?

 6.553s   mv /path/to/file /other/path/
 0.057s   cd /other/path/
19.088s   tar -xzf /other/path/file

I know that with time I can get the time spent of a single command. But I am looking for a solution for measuring the time of every command itself.

command-line bash time

2 Answers

  1. Best Answer

    You can use /usr/bin/time with the option -f like the following example. Preceed each command in your shellscript with /usr/bin/time -f "%E %C"

    $ /usr/bin/time -f "%E %C" sleep 2.34
0:02.34 sleep 2.34

    See man time for more details.

    Example:

    I made a small script, that can work to modify a simple shellscript by identifying commands, that can be tested with /usr/bin/time. Let us use the name time2script.

    #!/bin/bash

string='/usr/bin/time -f "%E %C"'

while IFS= read -r line
do
 cmd=$(<<< "$line" sed -e 's/ *//'|cut -d ' ' -f1)
 command=$(which "$cmd")
 if [ "$command" != "" ]
 then
  printf '%s %s\n' "$string" "$line"
 else
  printf '%s\n' "$line"
 fi
done < "$1"

    Using time2script on the example in the edited question:

    $ ./time2script socrates1
#!/bin/bash

/usr/bin/time -f "%E %C" mv /path/to/file /other/path/
cd /other/path/
/usr/bin/time -f "%E %C" tar -xzf /other/path/file

    Redirect to create a modified shellscript,

    ./time2script socrates1 > socrates1-timing
    • 6

  2. This is the naive approach. It has a few points of failure*, so I don't necessarily recommend using it, just wanted to mention it.

    #!/bin/bash
# Time each command in another Bash script.

script="$1"

# Skip blank lines and comments.
grep -vE "^\s*($|#)" "$script" |
    while IFS= read -r line
do
    echo "$line"
    time eval "$line"
done

    Example script foo.sh:

    #/bin/bash
true
sleep 2.34

    Example run:

    $ bash naive_timer.sh foo.sh
true

real    0m0.000s
user    0m0.000s
sys     0m0.000s
sleep 2.34

real    0m2.342s
user    0m0.000s
sys     0m0.000s

    * E.g. eval will choke on multi-line commands, like function declarations, array declarations, heredocs, even multi-line quotes. And commands that rely on the value of $0 may not work.

    • 2